From: Chitrak Gupta (chgupta_at_mix.wvu.edu)
Date: Fri Nov 20 2015 - 20:48:25 CST

Hi Christopher,

Thanks for the explanation.

I have another question. Is there a good way to generate IC tables for the
new molecule? Basically, transfer the information from the optimized PDB
into the *.top file? I am trying to mutate a wild-type protein and
introduce an unnatural amino acid. So I don't have an original PDB file for
the full protein. Because of this, PSFGEN is not being able to set the
coordinates for the new amino acid.

Regards,
Chitrak.

On Thu, Nov 19, 2015 at 4:23 PM, Mayne, Christopher G <cmayne2_at_illinois.edu>
wrote:

> Chitrak,
>
> I am still having a hard time getting the plots to match exactly. Just to
> clarify, it is the positions of the maxima/minima that matter the most,
> right? Not the actual value of the energy at each point (i.e. exact
> overlap)?
>
>
> To a large degree, both matter as you’d like to reproduce the conformation
> of the minimum energy and the general shape of the energy well. Getting an
> exact overlap, however, can be difficult depending on structural complexity
> and there are many situations in one wouldn’t expect MM calculations to
> match QM.
>
> Also, what does the "RMSE" in the fitting window indicate? Is it something
> like the mean-squared deviation, in that I should try to achieve as low a
> value as possible? Is there an approximate cutoff (something like, my RMSE
> must be below a certain number to be reliable)?
>
>
> RMSE is root mean square error of MM from QM, and is provided as a general
> metric for overall fit. Quickly looking over the code, the RMSE also takes
> weights into account, which means that portions of the PES that are above
> the energy cutoff (advanced settings) are excluded from the calculation.
> Because it is sensitive to the system and the weight settings, I can’t
> provide a good feeling for a reliable threshold…it will be system specific.
>
> Regards,
> Christopher Mayne
>
>
> On Nov 18, 2015, at 2:31 PM, Chitrak Gupta <chgupta_at_mix.wvu.edu> wrote:
>
> Hi Christopher,
>
> Thanks for the explanation. Actually, the roughness is more a result of me
> fiddling around with the dihedral parameters (the "r02" in the plot legend
> is misleading, I have done it multiple times, including writing to a
> temporary par file and starting over from those parameters). The first few
> refittings did not generate such a rough plot.
>
> I am still having a hard time getting the plots to match exactly. Just to
> clarify, it is the positions of the maxima/minima that matter the most,
> right? Not the actual value of the energy at each point (i.e. exact
> overlap)?
>
>
> Also, what does the "RMSE" in the fitting window indicate? Is it something
> like the mean-squared deviation, in that I should try to achieve as low a
> value as possible? Is there an approximate cutoff (something like, my RMSE
> must be below a certain number to be reliable)?
>
>
> Regards,
> Chitrak.
>
> On Wed, Nov 18, 2015 at 3:08 PM, Mayne, Christopher G <
> cmayne2_at_illinois.edu> wrote:
>
>> Chitrak,
>>
>> Those types of discontinuities are not uncommon. They are indicative of
>> major structural changes as the result of crossing some energy barrier
>> during the scan. The roughness of the MM PES, however, is a little odd. I
>> would recheck bonds and angles, if you parameterized any, as the dihedral
>> PES is often where trouble with earlier parameters will often show up.
>>
>> Regards,
>> Christopher Mayne
>>
>> > On Nov 18, 2015, at 8:07 AM, Chitrak Gupta <chgupta_at_mix.wvu.edu> wrote:
>> >
>> > Hi VMD users,
>> >
>> > I am trying to parameterize a unnatural amino acid using FFTK. I am at
>> the dihedral fitting stage. I only have one new dihedral to fit.
>> >
>> > What I find strange is that the QME plot has discontinuities. I ran the
>> Gaussian jobs with 5 degree step and 2 degree step, but both the times the
>> discontinuities persisted. Is this normal, or has someone else had this
>> issue as well?
>> >
>> > I have attached the QME plot here.
>> >
>> >
>> >
>> > Best regards,
>> > Chitrak.
>> > <Fitting.pdf>
>>
>>
>
>